> to use LORA fine-tuned model in inference you would still need the original model + trained additional layers, right? You don't need additional layers. Afte

> to use LORA fine-tuned model in inference you would still need the original model + trained additional layers, right? You don't need additional layers. After training, the product of the two matrices is added to the original weights matrix, so the model size remains the same as the original during inference.